On Tuesday, January 21, 2014 8:31:23 PM UTC-5, ywee...@gmail.com wrote:> Hi all, > > > > Sorry for the late reply as I went overseas for a couple of days, and didn't check the forum during this period. May I know if any one know any good papers on this technique, and can share with me its title, thanks! > > > > To John, > > Yes. The purpose is to detect the QPSK signal (either before demodulation or after demodulation; both way are allowed). Yes, that guy told me that there is this approach by raising the signal itself to the fourth power and this collapse the modulation into a carrier stick, and it will work for -20 dB SNR. > > > > To Mark: > > Yes, I mean -20dB, and it is in-band SNR meaning that the SNR is -20dB in the bassband of the QPSK signal. > > > > To julius: > > I am considering detecting the weak QPSK signal with a band of frequency from 500MHz to 700MHz, and I do not know exactly what frequency it is using. I am consider detecting the presence of a QPSK signal at a known frequency band.Well I still have my doubts about the viability of this, but I can suggest that you take a look at what comes out of your ^4 detector when there is no signal present, and look for differences against that baseline. John
Detection of weak QPSK signal, e.g., SNR=-20dB
Started by ●January 16, 2014
Reply by ●January 21, 20142014-01-21
Reply by ●January 21, 20142014-01-21
No reference signal or symbols or anything?
It's not impossible, but it will take a lot of work to calibrate the possible choices of {fc, R, \tau, \deltaF, doppler}.
I'll suggest this. Try to do detect such a signal in a synthetic simulation first, to keep it simple you can even assume that you know the exact symbols/bits that are being sent. To be able to do reliable detection, you have to process about 1000 symbols at a time, to get 30 dB of processing gain such that your SNReff = 10 dB. Now try to do the same without knowing the reference symbol. Then add perturbations as mentioned above.
High-order statistical methods are well-known and you need to do your own literature search. But each one of them assume a particular hypothesis {fc, R} and perhaps other pertubations. So overall the search space is complicated even for any given hypothesis.
Good luck.
Reply by ●January 22, 20142014-01-22
On Tue, 21 Jan 2014 18:40:42 -0800 (PST), julius <juliusk@gmail.com> wrote:>No reference signal or symbols or anything?=20They're not really necessary, since the nonlinearity removes the modulation, anyway.>It's not impossible, but it will take a lot of work to calibrate the possib= >le choices of {fc, R, \tau, \deltaF, doppler}.=20 > >I'll suggest this. Try to do detect such a signal in a synthetic simulation= > first, to keep it simple you can even assume that you know the exact symbo= >ls/bits that are being sent. To be able to do reliable detection, you have = >to process about 1000 symbols at a time, to get 30 dB of processing gain su= >ch that your SNReff =3D 10 dB. Now try to do the same without knowing the r= >eference symbol. Then add perturbations as mentioned above.=20 > >High-order statistical methods are well-known and you need to do your own l= >iterature search. But each one of them assume a particular hypothesis {fc, = >R} and perhaps other pertubations. So overall the search space is complicat= >ed even for any given hypothesis.=20 > >Good luck.Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com
Reply by ●January 22, 20142014-01-22
On Tue, 21 Jan 2014 17:31:23 -0800 (PST), yweesoon@gmail.com wrote:> >Hi all, > >Sorry for the late reply as I went overseas for a couple of days, and didn'= >t check the forum during this period. May I know if any one know any good p= >apers on this technique, and can share with me its title, thanks! > >To John, >Yes. The purpose is to detect the QPSK signal (either before demodulation o= >r after demodulation; both way are allowed). Yes, that guy told me that the= >re is this approach by raising the signal itself to the fourth power and th= >is collapse the modulation into a carrier stick, and it will work for -20 d= >B SNR.=20 > >To Mark: >Yes, I mean -20dB, and it is in-band SNR meaning that the SNR is -20dB in t= >he bassband of the QPSK signal. > >To julius: >I am considering detecting the weak QPSK signal with a band of frequency fr= >om 500MHz to 700MHz, and I do not know exactly what frequency it is using. = >I am consider detecting the presence of a QPSK signal at a known frequency = >band. > >FWIW, I've used this technique successfully in the past to detect and classify BPSK or QPSK signals and determine their frequency quickly. I'm a bit skeptical about the -20dB SNR part of it, depending on how you define SNR. There are some pitfalls with this technique, one of which is that it amplifies noise, so you have to be pretty careful about your constraints and what you're doing. I've seen this technique mentioned in a few papers, usually as a portion of a fast-acquisition strategy in a single-carrier communication system, but it's a pretty old and well-known method. e.g., it's mentioned here in the context of a method that can be implemented with analog/rf circuits: http://en.wikipedia.org/wiki/Carrier_recovery Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com
Reply by ●January 22, 20142014-01-22
On Wednesday, January 22, 2014 12:41:04 AM UTC-5, Eric Jacobsen wrote:> > They're not really necessary, since the nonlinearity removes the > modulation, anyway. >Yes, but detection of a known signal is always easier and provides a performance bound.
Reply by ●January 23, 20142014-01-23
On Wednesday, January 22, 2014 9:45:14 AM UTC-5, julius wrote:> On Wednesday, January 22, 2014 12:41:04 AM UTC-5, Eric Jacobsen wrote: > > They're not really necessary, since the nonlinearity removes the > modulation, anyway. > Yes, but detection of a known signal is always easier and provides a performance bound.To the OP you answered my question that the SNR you are asking about is indeed -20 dB. Next question...What excatly do you mena by "detection" Do you mean,demodulate and recover the data... or do you mean simply detect that a QPSK signal is present and not just random noise? Mark
Reply by ●January 24, 20142014-01-24
On Friday, January 24, 2014 1:23:22 AM UTC+8, mako...@yahoo.com wrote:> To the OP > > you answered my question that the SNR you are asking about is indeed -20 dB. > > Next question...What excatly do you mena by "detection" > > Do you mean,demodulate and recover the data... > or > > do you mean simply detect that a QPSK signal is present and not just random noise? > > MarkHi Mak: What I mean by detection is to detect the presence of the QPSK signal, and to know its carrier frequency. It does not neeed to recover the data. Thanks.
Reply by ●January 24, 20142014-01-24
yweesoon@gmail.com writes:> On Friday, January 24, 2014 1:23:22 AM UTC+8, mako...@yahoo.com wrote: >> To the OP >> >> you answered my question that the SNR you are asking about is indeed -20 dB. >> >> Next question...What excatly do you mena by "detection" >> >> Do you mean,demodulate and recover the data... >> or >> >> do you mean simply detect that a QPSK signal is present and not just random noise? >> >> Mark > > Hi Mak: > > What I mean by detection is to detect the presence of the QPSK signal, > and to know its carrier frequency. It does not neeed to recover the > data.Strictly speaking, the first is detection; the second is parameter estimation. -- Randy Yates Digital Signal Labs http://www.digitalsignallabs.com
Reply by ●January 24, 20142014-01-24
On Thursday, January 23, 2014 11:06:52 PM UTC-5, ywee...@gmail.com wrote:> On Friday, January 24, 2014 1:23:22 AM UTC+8, mako...@yahoo.com wrote: > To the OP > > you answered my question that the SNR you are asking about is indeed -20 dB. > > Next question...What excatly do you mena by "detection" > > Do you mean,demodulate and recover the data... > or > > do you mean simply detect that a QPSK signal is present and not just random noise? > > Mark Hi Mak: What I mean by detection is to detect the presence of the QPSK signal, and to know its carrier frequency. It does not neeed to recover the data. Thanks.==================== (I'm sorry if Google is doing a bad job of quoting) OK next question... What else do you already know "a priori" about the QPSK signal? for example do you already know the symbol rate? do you already know what data pattern it on it? anything else? Mark
Reply by ●January 26, 20142014-01-26
> OK next question...> > What else do you already know "a priori" about the QPSK signal? > for example > do you already know the symbol rate? > do you already know what data pattern it on it? > anything else? > > MarkThe only information known is the carrier frequency is from 500MHz to 700MHz, bit rate around 2Mbit/s, and it uses QPSK modulation. Thanks.
